Visual walkthrough — Pivot positions, free variables

Everything here elaborates Pivot positions, free variables. We lean on Row Reduction and Echelon Forms for the mechanics and point ahead to Null Space and Solution Sets of Ax=0 and Rank and the Rank-Nullity Theorem.

Step 1 — What a linear system even is (a grid of boxes)

WHAT. Before any symbols, picture the raw object. A system of linear equations in variables $x_1,x_2,x_3$ is just a table of numbers. Each row is one equation; each column holds all the coefficients that multiply one particular variable.

WHY. We must anchor the words "row" and "column" to a picture before we ever say "pivot", because a pivot lives at the crossing of a specific row and a specific column. If those two words are fuzzy, everything after is fuzzy.

PICTURE. In the figure, the orange band is one row (one equation). The teal band is one column (everything attached to the variable $x_2$). The number sitting where they cross — that single box — is the coefficient $a_{ij}$: row $i$, column $j$.

Nothing here is solved yet. We just named the parts.

Step 2 — Row reduction: staircase-ing the grid

WHAT. We now sweep the grid into a staircase shape (echelon form). Using the moves of Row Reduction and Echelon Forms — swap rows, scale a row, add a multiple of one row to another — we make the first nonzero number in each row sit strictly to the right of the one above it.

WHY. Why bother? Because a staircase is the shape in which each equation reveals exactly one variable it controls — the variable sitting at the tip of its step. That tip is the whole point of this page. A random grid hides who controls whom; the staircase makes it obvious.

PICTURE. Watch the plum "staircase edge" descend to the right. Each step's corner is the leading entry of that row — its first nonzero number after reduction.

$\underset{\text{first nonzero of row }i}{\underbrace{a_{ij}}}=\text{leading entry}\Rightarrow \text{column }j\text{ is a pivot column.}$

Step 3 — Each pivot pins exactly one variable

WHAT. Take one row of the staircase. Its leading entry sits in some column $j$. That row can be rearranged to read "$x_j$ equals (stuff made of the other variables)". So the pivot at column $j$ solves for $x_j$ — it pins that one variable down.

WHY. Why exactly one, never two? Because "leading" means first. Everything to the left of the leading entry in that row is already zero (that's what the staircase guarantees), so no earlier variable appears in this equation. The leading variable is the only one this equation is responsible for; all others are pushed to the right-hand side.

PICTURE. The orange arrow points from the pivot box to the variable it captures. Notice the greyed-out zero boxes to its left — that's why no earlier variable can steal the equation.

$\underset{\text{leading 1 in col }j}{\underbrace{1}}\cdot x_j+\underset{\text{later variables, moved right}}{\underbrace{c_{j+1}{x}_{j+1}+\cdots }}=d⟹x_j=d-(c_{j+1}{x}_{j+1}+\cdots )$

• The $1$ is the leading entry (already scaled to $1$ in RREF).
• $x_j$ — the pivot variable this row owns.
• Everything after it — later variables, which we shove to the right as "known if we're told them".

Step 4 — The leftover columns become free knobs

WHAT. Some columns never receive a leading entry — the staircase steps right past them. Look back at Step 2: the staircase skipped a column. No row is in charge of that variable. So nothing pins it. We are allowed to set it to any number we like.

WHY. Why call it "free"? Because the equations never constrain it. It is a knob: turn it to whatever value, and the pivot variables (Step 3) simply adjust to keep every equation true. This is the birth of a parameter.

PICTURE. The teal column has no staircase corner in it — the dial icon marks it as a free knob. Turn the dial and the orange pivot variables slide to compensate.

Step 5 — Counting: the columns split into exactly two piles

WHAT. Every column is one of two kinds, with no overlap and nothing left out: it either holds a pivot or it doesn't. So the total column count $n$ splits cleanly into (pivot columns) + (non-pivot columns).

WHY. Why is this a clean split? Because "has a pivot" is a yes/no question about a column — there's no third option. A column can't half-hold a pivot. This exhaustiveness is exactly what lets us subtract.

PICTURE. All $n$ columns lined up. Orange = pivot columns (there are $r$ of them). Teal = the rest. The two colours tile the whole strip with no gaps and no overlaps.

This is the parent note's headline result, now seen rather than asserted.

Step 6 — What the free count means geometrically

WHAT. The number of free knobs is the dimension of the solution set (when a solution exists). Zero knobs → a single point. One knob → a line. Two knobs → a plane.

WHY. Why does turning $k$ independent knobs sweep out a $k$-dimensional shape? Because each free variable is an independent direction you can slide along, and each independent slide adds one dimension — exactly the idea of Basis and Dimension. One slide traces a line; a second, independent slide fans the line out into a plane.

PICTURE. Three panels: $0$ knobs (a dot), $1$ knob (a line with an arrow $t\mathbf{v}$), $2$ knobs (a plane spanned by two arrows).

$\mathbf{x}=\underset{\text{one solution}}{\underbrace{\mathbf{p}}}+\underset{k\text{ free knobs}=k\text{ directions}}{\underbrace{t_1{\mathbf{v}}_1+\cdots +t_k{\mathbf{v}}_k}}$

The ${\mathbf{v}}_i$ are the directions the free variables push in — a basis for the null space when $\mathbf{b}=0$.

Step 7 — The degenerate cases we must not skip

WHAT. The formula $\text{free}=n-r$ has boundary behaviours. We check each so no reader hits a wall.

WHY. A rule you can't push to its extremes is a rule you don't trust. Here are all the edges.

PICTURE. Four mini-grids: (a) $r=n$ full rank, (b) zero rows present but full column rank, (c) more variables than usable equations, (d) an inconsistent augmented column.

• (a) $r=n$ (no free variables). Every column has a pivot; $\text{free}=n-n=0$. At most one solution — a point (Step 6, left panel).
• (b) Zero rows appear. A zero row does not create a free variable. Free variables come from non-pivot columns, never from empty rows. In the figure, a zero row sits at the bottom yet every column is still pivoted — $\text{free}=0$. This is the parent's first steel-manned mistake, drawn.
• (c) Wide matrix, $n>m$. With more variables than equations, at most $m$ pivots exist, so $r\le m<n$, forcing $\text{free}=n-r\ge 1$. There is always at least one free variable — a system with more unknowns than equations can never pin everything.
• (d) Inconsistent. If the augmented column $[\mathbf{b}]$ itself holds a pivot — a row reading $[0\cdots 0\mid c]$ with $c\ne 0$ — that's the false statement $0=c$. The solution set is empty, and the free count describes nothing. Always check consistency before you count knobs.

The one-picture summary

This final figure folds all seven steps into one flow: raw grid → staircase → pivots pin variables → leftover columns float free → count them as $n-r$ → read off the solution's dimension, guarded by a consistency check.

Recall checkpoints

Connections

• Pivot positions, free variables — parent topic
• Row Reduction and Echelon Forms
• Rank and the Rank-Nullity Theorem
• Null Space and Solution Sets of Ax=0
• Linear Independence
• Existence and Uniqueness of Solutions
• Basis and Dimension

Concept Map